Re: [idn] Where to do form-folding?

[James Seng <James@Seng.cc>]

Dan Oscarsson wrote:
> Right. The data on the wire will be different for the two cases.
> In one the queried named will be form-folded, in the other all
> character data will have the same format.

Bingo! We should discuss if canonicalization should occur before it hits
the wire or after, but lets not go into whether it should be in the
library or in the application.

After spending some time thinking (actually spending more time taking
photo with Pikachu..but thats another story), it seem both method will
work but which is more 'clean'. And IMHO, the client-based solution seem
to be better.

"J. William Semich" wrote:
> My view: "form-folding" will be required on the application level, yes. but
> it is not for certain all applications will do it, or do it
> appropriately/authoritatively.

If it does not do it appropriately, then it is not following whatever
standard, hence it is broken.

I believe Yoneya-san has an interesting way to have authoritative
canonicalization and doing it consistently. Hence, it is not as if it is
impossible to do.

> The question between 1 and 2 below is, where will the "authoritative"
> folding/canonicalization take place? IMO, DNS servers should be the
> authoritative point of reference, as they will continue to be for ASCII
> case-insensitivity. Implementation tests can show how much impact this will
> have on CPU resources.

When you have 10,000 queries/sec, CPU resource will be a concern.

-James Seng